GCE LoadBalancer IP is pending indefinitely


#1

I’ve followed along several of the blog posts in the “A Service Mesh for Kubernetes” series. Each time, the LoadBalancer is not created and the IP is pending indefinitely.

Maybe I am missing a service account / the correct permissions on the VMs? (Not GKE, Ubuntu 16.04 VMs with k8s 1.6.4)


#2

Is there anything in Flavors of Kubernetes that is helpful?


#3

Am not an expert on this by any means, but if you’re running in GCE, not GKE, then I think you need to configure your cluster with -cloud-provider=gce in order for it to automatically provision external IP addresses.


#4

@verdverm did Kevin’s suggestion help?


#5

Hello,

I’m running into this same issue in a GCE cluster. I’m running K8s 1.9, I’ve set --cloud-provider=gce on both the kubelets and the kube-apiserver but load balancer provisioning gets stuck in . Any tips you can provide to debug this? Do I need to put --cloud-provider=gce on any other compontents?

Best,

Claude


#6

Hey @claudenm – it would be worth determining if this is a linkerd-specific issue, or if your cluster is incapable of provisioning external IPs for any service. I recommend walking through this guide to figure it:

https://kubernetes.io/docs/tutorials/stateless-application/expose-external-ip-address/


#7

@klingerf – Thanks. Between the time I asked the question and your response I got the logs for the l5d pods and saw I was out of Static IPs (over quota in GCE). Embarrassing for me :blush: . Thanks for your help!


#8

@claudenm Glad you got it figured out!